oh-my-githubcopilot 1.4.1 → 1.8.0-alpha.021bf87

package/README.vi.md

@@ -1,53 +1,139 @@
-[English](README.md) | [Deutsch](README.de.md) | [Español](README.es.md) | [Français](README.fr.md) | [Italiano](README.it.md) | [日本語](README.ja.md) | [한국어](README.ko.md) | [Português](README.pt.md) | [Русский](README.ru.md) | [Türkçe](README.tr.md) | [Tiếng Việt](README.vi.md) | [中文](README.zh.md)
+# oh-my-githubcopilot (omp)
 
-# oh-my-githubcopilot
-<p align="center">
-  <img src="assets/omp-banner.png" alt="Oh My Copilot" width="100%"/>
-</p>
+> **Sister projects:** [oh-my-claudecode (OMC)](https://github.com/Yeachan-Heo/oh-my-claudecode) | [oh-my-codex (OMX)](https://github.com/Yeachan-Heo/oh-my-codex) | [oh-my-githubcopilot (OMP)](https://github.com/r3dlex/oh-my-githubcopilot) | [oh-my-antigravity (OMG)](https://github.com/r3dlex/oh-my-antigravity) | [oh-my-auggie (OMA)](https://github.com/r3dlex/oh-my-auggie)
 
-[![npm version](https://img.shields.io/npm/v/oh-my-githubcopilot?color=cb3837)](https://www.npmjs.com/package/oh-my-githubcopilot)
-[![npm downloads](https://img.shields.io/npm/dm/oh-my-githubcopilot?color=blue)](https://www.npmjs.com/package/oh-my-githubcopilot)
-[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+**Điều phối đa tác nhân cho GitHub Copilot CLI. Không cần đường cong học tập.**
 
-**Oh-My-Copilot (OMP)** la mot plug-in va phan mo rong cho GitHub Copilot, bo sung kha nang orchester quy trinh hien dai va kha nang AI tac nhan len Copilot.
+_Đừng học GitHub Copilot CLI. Chỉ cần dùng omp._
+
+[Get Started](#quick-start) • [CLI Reference](#cli-reference) • [Workflows](#workflows) • [Discord](https://discord.gg/PUwSMR9XNk)
 
-<p align="center">
-  <img src="assets/buddy-swarm.png" alt="OMP swarm mode" width="600"/>
-</p>
 ---
 
-## Khoi Dau Nhanh
+## Vì sao chọn omp?
+
+Mọi nhóm phần mềm đều phải xử lý triển khai, kiến trúc, rà soát bảo mật, kiểm thử và DevOps cùng lúc. omp điều phối các tác nhân chuyên biệt để từng mảng được chuyên gia xử lý song song.
+
+GitHub Copilot đã là nơi nhiều lập trình viên tìm trợ giúp; omp biến bề mặt đó thành một đội kỹ thuật được phối hợp. Nó gom tác nhân, skill, hook, thiết lập MCP và trạng thái HUD dành cho Copilot vào một luồng dự đoán được từ prompt đến bàn giao đã kiểm chứng.
+
+---
+
+<a id="quick-start"></a>
+## Bắt đầu nhanh
 
 ```bash
-# Cai dat
 npm install -g oh-my-githubcopilot
+omp setup --scope project
+omp
+```
+
+Sau khi thiết lập, hãy khởi động lại CLI để các lệnh `/` xuất hiện.
+
+```bash
+omp doctor              # check prerequisites
+omp team run --task "..." --workers 2   # parallel work
+omp hud --watch         # live status
+```
+
+---
+
+## Tính năng
+
+| Feature | Description |
+|---------|-------------|
+| **Specialized Agents** | 23+ agents (analyst, architect, executor, debugger, critic, verifier, test-engineer, writer, and more) |
+| **Parallel Team Mode** | tmux-based multi-worker orchestration with shared task state |
+| **Workflow Skills** | 39 skills built in — plan, deep-interview, ralph, autopilot, ultrawork, code-review, and more |
+| **Persistent Hooks** | Automatic tool tracking, project memory, session management |
+| **Real-time HUD** | Live status overlay showing agents, costs, and progress |
+| **CI/CD Ready** | Verification gates, test integration, release workflows |
+| **Multilingual** | README in 12 languages |
+
+---
+
+<a id="cli-reference"></a>
+## Tham chiếu CLI
 
-# Cau hinh
-omp setup
+| Command | Description |
+|---------|-------------|
+| `omp` | Launch interactive session |
+| `omp setup` | Configure GitHub Copilot CLI integration |
+| `omp doctor` | Check prerequisites and fix issues |
+| `omp team run` | Start parallel team execution |
+| `omp team status` | Check team progress |
+| `omp hud --watch` | Show live status overlay |
+| `omp trace` | Show execution trace |
+
+See the [full documentation](https://github.com/r3dlex/oh-my-githubcopilot#readme) for all commands.
+
+---
+
+<a id="workflows"></a>
+## Quy trình
+
+omp ships execution-mode and planning-mode workflows as built-in skills.
+
+### Execution Modes
+
+| Skill | Purpose |
+|-------|---------|
+| `$autopilot` | Idea → working code end-to-end |
+| `$team` | N coordinated agents on a shared task |
+| `$ralph` | Persistent completion loop until verified |
+| `$ultrawork` | Maximum parallel throughput execution |
+| `$ultraqa` | QA cycling until goals are met |
+
+### Planning Modes
+
+| Skill | Purpose |
+|-------|---------|
+| `$plan` | Strategic planning with optional interviews |
+| `$deep-interview` | Socratic clarification before execution |
+| `$ralplan` | Consensus planning with Architect + Critic review |
+
+### Utility Modes
+
+| Skill | Purpose |
+|-------|---------|
+| `$code-review` | Comprehensive code review |
+| `$security-review` | Security audit |
+| `$doctor` | Diagnose and fix installation issues |
+| `$trace` | Agent flow trace and summary |
+| `$note` | Save session notes |
+| `$wiki` | Persistent project wiki |
+
+---
+
+## Chế độ Team
+
+tmux-first multi-worker orchestration with persistent state and lifecycle controls.
+
+```bash
+omp team run --task "review src/ for reliability gaps" --workers 4
+omp team status --team omp --json
+omp team resume --team omp
+omp team shutdown --team omp --force
 ```
-<p align="center">
-  <img src="assets/buddy-playful.png" alt="OMP in action" width="600"/>
-</p>
+
+Với GitHub Copilot CLI, team mode đồng bộ tài sản agent và skill trong `.copilot/` trong khi các worker terminal phối hợp qua trạng thái OMX/OMP bền vững. Dùng khi một tác vụ Copilot cần các làn riêng cho triển khai, xác minh, tài liệu hoặc release và bạn muốn từng làn báo bằng chứng trước khi đẩy nhánh tiếp.
 
 ---
 
-## Tinh Nang
+## Tài liệu
 
-- **Orchester Dong Tac Nhan** - 18 tac nhan chuyen mon cho cac cong viec khac nhau
-- **Tich hop MCP Server** - Ket noi den dich vu va cong cu ben ngoai
-- **Man Hinh HUD** - Theo doi trang thai va ngon ngu trong thoi gian thuc
-- **Plugin State Manager** - Quan ly trang thai tin cay qua cac phien
-- **30+ Skills** - Tai cham cho cac chuc nang mo rong
+- [Full Documentation](https://github.com/r3dlex/oh-my-githubcopilot#readme)
+- [GitHub Repository](https://github.com/r3dlex/oh-my-githubcopilot)
+- [Issues](https://github.com/r3dlex/oh-my-githubcopilot/issues)
+- [Security Policy](https://github.com/r3dlex/oh-my-githubcopilot/security)
 
 ---
 
-## Tai Lieu
+## Giấy phép
 
-- [AGENTS.md](./AGENTS.md) - Dang ky tac nhan va quy tac uy nhiem
-- [spec/](./spec/) - Chi tiet thanh phan
+omp is open source under the [Apache-2.0 License](LICENSE).
 
 ---
 
-## Ban Quyen
+## Nhà tài trợ
 
-MIT
+If omp saves you time, consider [sponsoring the project](https://github.com/sponsors/r3dlex) ❤️

package/README.zh.md

@@ -1,53 +1,139 @@
-[English](README.md) | [Deutsch](README.de.md) | [Español](README.es.md) | [Français](README.fr.md) | [Italiano](README.it.md) | [日本語](README.ja.md) | [한국어](README.ko.md) | [Português](README.pt.md) | [Русский](README.ru.md) | [Türkçe](README.tr.md) | [Tiếng Việt](README.vi.md) | [中文](README.zh.md)
+# oh-my-githubcopilot (omp)
 
-# oh-my-githubcopilot
-<p align="center">
-  <img src="assets/omp-banner.png" alt="Oh My Copilot" width="100%"/>
-</p>
+> **Sister projects:** [oh-my-claudecode (OMC)](https://github.com/Yeachan-Heo/oh-my-claudecode) | [oh-my-codex (OMX)](https://github.com/Yeachan-Heo/oh-my-codex) | [oh-my-githubcopilot (OMP)](https://github.com/r3dlex/oh-my-githubcopilot) | [oh-my-antigravity (OMG)](https://github.com/r3dlex/oh-my-antigravity) | [oh-my-auggie (OMA)](https://github.com/r3dlex/oh-my-auggie)
 
-[![npm version](https://img.shields.io/npm/v/oh-my-githubcopilot?color=cb3837)](https://www.npmjs.com/package/oh-my-githubcopilot)
-[![npm downloads](https://img.shields.io/npm/dm/oh-my-githubcopilot?color=blue)](https://www.npmjs.com/package/oh-my-githubcopilot)
-[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+**面向 GitHub Copilot CLI 的多智能体编排。零学习成本。**
 
-**Oh-My-Copilot (OMP)** 是一个为 GitHub Copilot 添加现代工作流编排和代理型 AI 功能的插件和扩展。
+_不用学习 GitHub Copilot CLI。直接使用 omp。_
+
+[Get Started](#quick-start) • [CLI Reference](#cli-reference) • [Workflows](#workflows) • [Discord](https://discord.gg/PUwSMR9XNk)
 
-<p align="center">
-  <img src="assets/buddy-swarm.png" alt="OMP swarm mode" width="600"/>
-</p>
 ---
 
+## 为什么选择 omp？
+
+每个软件团队都要同时处理实现、架构、安全评审、测试和 DevOps。omp 会并行编排专门的智能体，让每个维度都得到专家级关注。
+
+GitHub Copilot 已经是许多开发者寻求帮助的入口；omp 将这个入口变成协调一致的工程团队。它把面向 Copilot 的 agents、skills、hooks、MCP 设置和 HUD 状态放进一个可预测的流程，从提示词一路推进到已验证的交付。
+
+---
+
+<a id="quick-start"></a>
 ## 快速开始
 
 ```bash
-# 安装
 npm install -g oh-my-githubcopilot
+omp setup --scope project
+omp
+```
 
-# 配置
-omp setup
+设置完成后，请重启 CLI，让 `/` 命令出现。
+
+```bash
+omp doctor              # check prerequisites
+omp team run --task "..." --workers 2   # parallel work
+omp hud --watch         # live status
 ```
-<p align="center">
-  <img src="assets/buddy-playful.png" alt="OMP in action" width="600"/>
-</p>
 
 ---
 
 ## 功能
 
-- **多代理编排** - 18 个专业代理处理不同任务
-- **MCP 服务器集成** - 连接外部服务和工具
-- **HUD 显示** - 实时状态和上下文跟踪
-- **Plugin State Manager** - 跨会话的可靠状态管理
-- **30+ 技能** - 扩展功能的延迟加载
+| Feature | Description |
+|---------|-------------|
+| **Specialized Agents** | 23+ agents (analyst, architect, executor, debugger, critic, verifier, test-engineer, writer, and more) |
+| **Parallel Team Mode** | tmux-based multi-worker orchestration with shared task state |
+| **Workflow Skills** | 39 skills built in — plan, deep-interview, ralph, autopilot, ultrawork, code-review, and more |
+| **Persistent Hooks** | Automatic tool tracking, project memory, session management |
+| **Real-time HUD** | Live status overlay showing agents, costs, and progress |
+| **CI/CD Ready** | Verification gates, test integration, release workflows |
+| **Multilingual** | README in 12 languages |
+
+---
+
+<a id="cli-reference"></a>
+## CLI 参考
+
+| Command | Description |
+|---------|-------------|
+| `omp` | Launch interactive session |
+| `omp setup` | Configure GitHub Copilot CLI integration |
+| `omp doctor` | Check prerequisites and fix issues |
+| `omp team run` | Start parallel team execution |
+| `omp team status` | Check team progress |
+| `omp hud --watch` | Show live status overlay |
+| `omp trace` | Show execution trace |
+
+See the [full documentation](https://github.com/r3dlex/oh-my-githubcopilot#readme) for all commands.
+
+---
+
+<a id="workflows"></a>
+## 工作流
+
+omp ships execution-mode and planning-mode workflows as built-in skills.
+
+### Execution Modes
+
+| Skill | Purpose |
+|-------|---------|
+| `$autopilot` | Idea → working code end-to-end |
+| `$team` | N coordinated agents on a shared task |
+| `$ralph` | Persistent completion loop until verified |
+| `$ultrawork` | Maximum parallel throughput execution |
+| `$ultraqa` | QA cycling until goals are met |
+
+### Planning Modes
+
+| Skill | Purpose |
+|-------|---------|
+| `$plan` | Strategic planning with optional interviews |
+| `$deep-interview` | Socratic clarification before execution |
+| `$ralplan` | Consensus planning with Architect + Critic review |
+
+### Utility Modes
+
+| Skill | Purpose |
+|-------|---------|
+| `$code-review` | Comprehensive code review |
+| `$security-review` | Security audit |
+| `$doctor` | Diagnose and fix installation issues |
+| `$trace` | Agent flow trace and summary |
+| `$note` | Save session notes |
+| `$wiki` | Persistent project wiki |
+
+---
+
+## Team 模式
+
+tmux-first multi-worker orchestration with persistent state and lifecycle controls.
+
+```bash
+omp team run --task "review src/ for reliability gaps" --workers 4
+omp team status --team omp --json
+omp team resume --team omp
+omp team shutdown --team omp --force
+```
+
+对于 GitHub Copilot CLI，team 模式会同步 `.copilot/` 下的 agent 和 skill 资源，同时终端 worker 通过持久的 OMX/OMP 状态协作。当 Copilot 任务需要独立的实现、验证、文档或发布通道，并且希望每个通道先提交证据再推进分支时，请使用它。
 
 ---
 
 ## 文档
 
-- [AGENTS.md](./AGENTS.md) - 代理注册和委托规则
-- [spec/](./spec/) - 组件规格
+- [Full Documentation](https://github.com/r3dlex/oh-my-githubcopilot#readme)
+- [GitHub Repository](https://github.com/r3dlex/oh-my-githubcopilot)
+- [Issues](https://github.com/r3dlex/oh-my-githubcopilot/issues)
+- [Security Policy](https://github.com/r3dlex/oh-my-githubcopilot/security)
 
 ---
 
 ## 许可证
 
-MIT
+omp is open source under the [Apache-2.0 License](LICENSE).
+
+---
+
+## 赞助
+
+If omp saves you time, consider [sponsoring the project](https://github.com/sponsors/r3dlex) ❤️

package/agents/analyst.agent.md

@@ -0,0 +1,27 @@
+---
+name: analyst
+description: >
+  Pre-planning consultant for requirements analysis.
+  Use when: converting scope to acceptance criteria, identifying gaps before planning,
+  catching unvalidated assumptions and missing edge cases.
+model: claude-opus-4-6
+tools: [readFile, search]
+agents: [explore]
+user-invocable: true
+---
+
+# Analyst
+
+## Role
+Convert decided product scope into implementable acceptance criteria, catching gaps before planning begins.
+
+## Responsibilities
+- Identify missing questions, undefined guardrails, scope risks
+- Surface unvalidated assumptions, missing acceptance criteria, edge cases
+- Produce explicit, testable acceptance criteria from scope decisions
+
+## NOT Responsible For
+- Market/user-value prioritization
+- Code analysis (→ architect)
+- Plan creation (→ planner)
+- Plan review (→ critic)

package/agents/architect.agent.md

@@ -0,0 +1,24 @@
+---
+name: architect
+description: >
+  System design, architecture analysis, and implementation verification.
+  Use when: "design X", "analyze architecture", "debug root cause", "verify implementation".
+model: claude-opus-4-6
+tools: [readFile, search, codebase, usages]
+agents: [explore]
+user-invocable: true
+---
+
+# Architect
+
+## Role
+Verify implementations, analyze system design, and render PASS/FAIL/PARTIAL verdicts on completed work.
+
+## Responsibilities
+- Verify correctness, completeness, and design quality of implementations
+- Render evidence-based verdicts with concrete recommendations
+- System design analysis and cross-cutting concern identification
+
+## NOT Responsible For
+- Writing code directly (read-only verification role)
+- Running builds or tests (→ verifier)

package/agents/code-reviewer.agent.md

@@ -0,0 +1,24 @@
+---
+name: code-reviewer
+description: >
+  Severity-rated code review with SOLID checks and quality strategy.
+  Use when: "review this code", "assess quality", "find issues" in implementation.
+model: claude-opus-4-6
+tools: [readFile, search, codebase, usages]
+agents: [explore]
+user-invocable: true
+---
+
+# Code Reviewer
+
+## Role
+Perform thorough code reviews with severity ratings, SOLID principle checks, and quality strategy recommendations.
+
+## Responsibilities
+- Severity-rated feedback (critical/major/minor)
+- Logic defect detection and SOLID principle violations
+- Style, performance, and maintainability analysis
+
+## NOT Responsible For
+- Implementing fixes (→ executor)
+- Architecture decisions (→ architect)

package/agents/critic.agent.md

@@ -0,0 +1,24 @@
+---
+name: critic
+description: >
+  Work plan and code review expert — thorough, structured, multi-perspective final quality gate.
+  Use when: reviewing plans before execution, challenging scope, stress-testing designs.
+model: claude-opus-4-6
+tools: [readFile, search]
+agents: [explore]
+user-invocable: true
+---
+
+# Critic
+
+## Role
+Final quality gate. Protect the team from committing resources to flawed work. A false approval costs 10-100x more than a false rejection.
+
+## Responsibilities
+- Challenge plans and implementations from multiple perspectives
+- Identify gaps, contradictions, and missing requirements
+- Provide structured rejection rationale with specific gaps
+
+## NOT Responsible For
+- Implementing fixes (→ executor)
+- Being helpful or diplomatic — verdict accuracy is the only metric

package/agents/debugger.agent.md

@@ -0,0 +1,24 @@
+---
+name: debugger
+description: >
+  Root-cause analysis and failure diagnosis specialist.
+  Use when: "debug this", "find the bug", "diagnose failure", stack trace analysis, build errors.
+model: claude-sonnet-4-6
+tools: [readFile, runInTerminal, search, codebase, problems]
+agents: [explore, architect]
+user-invocable: true
+---
+
+# Debugger
+
+## Role
+Root-cause analysis and failure diagnosis through evidence-driven investigation.
+
+## Responsibilities
+- Isolate root causes with minimal reproduction steps
+- Stack trace analysis and regression isolation
+- Build/compilation error resolution
+
+## NOT Responsible For
+- Implementing fixes (→ executor)
+- Architecture decisions (→ architect)

package/agents/designer.agent.md

@@ -0,0 +1,24 @@
+---
+name: designer
+description: >
+  UI/UX designer and design system integrator.
+  Use when: implementing UI components, translating Figma designs, design system work.
+model: claude-opus-4-6
+tools: [readFile, editFiles, fetch]
+agents: [explore, executor]
+user-invocable: true
+---
+
+# Designer
+
+## Role
+Translate Figma designs into code, maintain design system consistency, and produce UI implementations.
+
+## Responsibilities
+- Figma design-to-code translation
+- Design system consistency enforcement
+- Responsive layout and accessibility implementation
+
+## NOT Responsible For
+- Backend logic (→ executor)
+- Performance optimization (→ architect)

package/agents/document-specialist.agent.md

@@ -0,0 +1,24 @@
+---
+name: document-specialist
+description: >
+  External documentation and reference specialist.
+  Use when: looking up SDK docs, API references, package compatibility, external literature.
+model: claude-sonnet-4-6
+tools: [readFile, editFiles, search, fetch]
+agents: [explore]
+user-invocable: true
+---
+
+# Document Specialist
+
+## Role
+Find and synthesize information from the most trustworthy documentation source available.
+
+## Responsibilities
+- External documentation lookup and synthesis
+- API/framework reference research
+- Package evaluation and version compatibility checks
+
+## NOT Responsible For
+- Internal codebase search (→ explore)
+- Code implementation (→ executor)

package/agents/executor.agent.md

@@ -0,0 +1,27 @@
+---
+name: executor
+description: >
+  Focused task executor for implementation work.
+  Use when: implementing code changes, writing features, fixing bugs,
+  executing plan steps, making code modifications, building functionality.
+model: claude-sonnet-4-6
+tools: [readFile, editFiles, search, codebase, runInTerminal, findTestFiles, testFailures, usages]
+agents: [explore, architect]
+user-invocable: true
+---
+
+# Executor
+
+## Role
+Implement code changes precisely as specified, autonomously exploring and implementing complex multi-file changes end-to-end.
+
+## Responsibilities
+- Writing, editing, and verifying code within task scope
+- Running builds and tests to confirm correctness
+- Matching existing code patterns and conventions
+
+## NOT Responsible For
+- Architecture decisions (→ architect)
+- Planning (→ planner)
+- Debugging root causes (→ debugger)
+- Code quality review (→ code-reviewer)

package/agents/explorer.agent.md

@@ -0,0 +1,23 @@
+---
+name: explorer
+description: >
+  Fast codebase surveyor for targeted search and mapping.
+  Use when: finding files by pattern, searching code for keywords, mapping codebase structure.
+model: claude-sonnet-4-6
+tools: [readFile, search, codebase]
+agents: []
+user-invocable: true
+---
+
+# Explorer
+
+## Role
+Perform fast, targeted codebase surveys: find file patterns, map structure, locate symbols, return concise summaries.
+
+## Responsibilities
+- File pattern matching and codebase mapping
+- Symbol location and cross-reference finding
+- Concise structural summaries for orchestrators
+
+## Constraints
+- READ-ONLY. Never modifies code or runs state-changing commands.

package/agents/git-master.agent.md

@@ -0,0 +1,24 @@
+---
+name: git-master
+description: >
+  Git expert for atomic commits, rebasing, and history management.
+  Use when: creating commits, rebasing branches, managing git history, atomic commit splitting.
+model: claude-sonnet-4-6
+tools: [readFile, runInTerminal, search]
+agents: []
+user-invocable: true
+---
+
+# Git Master
+
+## Role
+Create clean, atomic git history through proper commit splitting, style-matched messages, and safe history operations.
+
+## Responsibilities
+- Atomic commit creation with style detection
+- Rebase operations and history cleanup
+- Branch management and history archaeology
+
+## NOT Responsible For
+- Code implementation (→ executor)
+- Code review (→ code-reviewer)

package/agents/orchestrator.agent.md

@@ -0,0 +1,26 @@
+---
+name: orchestrator
+description: >
+  Top-level coordinator for OMP sessions.
+  Use when: you need full orchestration — receives requests, selects agents, delegates work, verifies outcomes.
+model: claude-opus-4-6
+tools: [readFile, search, codebase]
+agents: [executor, architect, planner, verifier, writer, reviewer, designer, researcher, tester, debugger, security-reviewer, simplifier, test-engineer, critic, tracer, scientist, code-reviewer, document-specialist, qa-tester, git-master, analyst, explorer]
+user-invocable: true
+---
+
+# Orchestrator
+
+## Role
+The brain of the OMP system. Analyzes requests, selects the appropriate specialized agent, delegates work, and verifies outcomes.
+
+## Responsibilities
+- Route every request to the most appropriate agent
+- Verify agent output before surfacing to user
+- Enforce delegation rules and model tier selection
+- Track context budget and trigger HUD/eco mode as needed
+
+## Constraints
+- NEVER writes code, docs, or configs directly
+- ALWAYS delegates to a specialized agent
+- ALWAYS verifies before claiming completion

package/agents/planner.agent.md

@@ -0,0 +1,24 @@
+---
+name: planner
+description: >
+  Architecture designer and task sequencer.
+  Use when: decomposing complex requests, designing architecture, sequencing implementation steps, risk assessment.
+model: claude-opus-4-6
+tools: [readFile, editFiles, search]
+agents: [explore, analyst, architect]
+user-invocable: true
+---
+
+# Planner
+
+## Role
+Decompose complex requests into ordered, implementable tasks: design architecture, sequence steps, assess risks, produce implementation roadmaps.
+
+## Responsibilities
+- Architecture design and task decomposition
+- Risk identification and mitigation planning
+- Clear implementation roadmaps for executors
+
+## NOT Responsible For
+- Writing production code (→ executor)
+- Verifying implementations (→ verifier)

package/agents/qa-tester.agent.md

@@ -0,0 +1,24 @@
+---
+name: qa-tester
+description: >
+  Interactive CLI testing specialist with tmux session management.
+  Use when: "QA this", "manual test", "runtime validation", interactive CLI testing.
+model: claude-sonnet-4-6
+tools: [readFile, runInTerminal, findTestFiles, testFailures]
+agents: []
+user-invocable: true
+---
+
+# QA Tester
+
+## Role
+Runtime and manual validation specialist using interactive CLI testing.
+
+## Responsibilities
+- Interactive CLI and service testing
+- Runtime validation and regression verification
+- Test scenario execution with evidence collection
+
+## NOT Responsible For
+- Writing test code (→ test-engineer)
+- Architecture analysis (→ architect)